Harald Piringer

Uncertainty-aware exploration of continuous parameter spaces using multivariate prediction

Systems projecting a continuous n‐dimensional parameter space to a continuous m‐dimensional target space play an important role in science and engineering. If evaluating the system is expensive, however, an analysis is often limited to a small number of sample points. The main contribution of this paper is an interactive approach to enable a continuous analysis of a sampled parameter space with respect to multiple target values. We employ methods from statistical learning to predict results in real‐time at any user‐defined point and its neighborhood. In particular, we describe techniques to guide the user to potentially interesting parameter regions, and we visualize the inherent uncertainty of predictions in 2D scatterplots and parallel coordinates. An evaluation describes a real‐world scenario in the application context of car engine design and reports feedback of domain experts. The results indicate that our approach is suitable to accelerate a local sensitivity analysis of multiple target dimensions, and to determine a sufficient local sampling density for interesting parameter regions.

Visual Parameter Space Analysis: A Conceptual Framework

Various case studies in different application domains have shown the great potential of visual parameter space analysis to support validating and using simulation models. In order to guide and systematize research endeavors in this area, we provide a conceptual framework for visual parameter space analysis problems. The framework is based on our own experience and a structured analysis of the visualization literature. It contains three major components: (1) a data flow model that helps to abstractly describe visual parameter space analysis problems independent of their application domain; (2) a set of four navigation strategies of how parameter space analysis can be supported by visualization tools; and (3) a characterization of six analysis tasks. Based on our framework, we analyze and classify the current body of literature, and identify three open research gaps in visual parameter space analysis. The framework and its discussion are meant to support visualization designers and researchers in characterizing parameter space analysis problems and to guide their design and evaluation processes.

Interactive focus+context visualization with linked 2D/3D scatterplots

Scatterplots in 2D and 3D are very useful tools, but also suffer from a number of problems. Overplotting hides the true number of points that are displayed, and showing point clouds in 3D is problematic both in terms of perception and interaction. We propose a combination of 2D and 3D scatterplots, together with some extensions to overcome these problems. By linking 2D and 3D views, it is possible to interact in 2D and get feedback in 3D. That feedback is also enhanced by depth cues (color and point size) such that the user gets a better depth impression. Histograms in 2D and 3D show additional information about point densities and additional context can be displayed. An example application demonstrates the usefulness of the technique.

Hypermoval: interactive visual validation of regression models for real-time simulation

During the development of car engines, regression models that are based on machine learning techniques are increasingly important for tasks which require a prediction of results in real‐time. While the validation of a model is a key part of its identification process, existing computation‐ or visualization‐based techniques do not adequately support all aspects of model validation. The main contribution of this paper is an interactive approach called HyperMoVal that is designed to support multiple tasks related to model validation: 1) comparing known and predicted results, 2) analyzing regions with a bad fit, 3) assessing the physical plausibility of models also outside regions covered by validation data, and 4) comparing multiple models. The key idea is to visually relate one or more n‐dimensional scalar functions to known validation data within a combined visualization. HyperMoVal lays out multiple 2D and 3D sub‐projections of the n‐dimensional function space around a focal point. We describe how linking HyperMoVal to other views further extends the possibilities for model validation. Based on this integration, we discuss steps towards supporting the entire workflow of identifying regression models. An evaluation illustrates a typical workflow in the application context of car‐engine design and reports general feedback of domain experts and users of our approach. These results indicate that our approach significantly accelerates the identification of regression models and increases the confidence in the overall engineering process.

A Partition-Based Framework for Building and Validating Regression Models

Regression models play a key role in many application domains for analyzing or predicting a quantitative dependent variable based on one or more independent variables. Automated approaches for building regression models are typically limited with respect to incorporating domain knowledge in the process of selecting input variables (also known as feature subset selection). Other limitations include the identification of local structures, transformations, and interactions between variables. The contribution of this paper is a framework for building regression models addressing these limitations. The framework combines a qualitative analysis of relationship structures by visualization and a quantification of relevance for ranking any number of features and pairs of features which may be categorical or continuous. A central aspect is the local approximation of the conditional target distribution by partitioning 1D and 2D feature domains into disjoint regions. This enables a visual investigation of local patterns and largely avoids structural assumptions for the quantitative ranking. We describe how the framework supports different tasks in model building (e.g., validation and comparison), and we present an interactive workflow for feature subset selection. A real-world case study illustrates the step-wise identification of a five-dimensional model for natural gas consumption. We also report feedback from domain experts after two months of deployment in the energy sector, indicating a significant effort reduction for building and improving regression models.

A Multi-Threading Architecture to Support Interactive Visual Exploration

During continuous user interaction, it is hard to provide rich visual feedback at interactive rates for datasets containing millions of entries. The contribution of this paper is a generic architecture that ensures responsiveness of the application even when dealing with large data and that is applicable to most types of information visualizations. Our architecture builds on the separation of the main application thread and the visualization thread, which can be cancelled early due to user interaction. In combination with a layer mechanism, our architecture facilitates generating previews incrementally to provide rich visual feedback quickly. To help avoiding common pitfalls of multi-threading, we discuss synchronization and communication in detail. We explicitly denote design choices to control trade-offs. A quantitative evaluation based on the system VI S P L ORE shows fast visual feedback during continuous interaction even for millions of entries. We describe instantiations of our architecture in additional tools.

Opening the Black Box: Strategies for Increased User Involvement in Existing Algorithm Implementations

An increasing number of interactive visualization tools stress the integration with computational software like MATLAB and R to access a variety of proven algorithms. In many cases, however, the algorithms are used as black boxes that run to completion in isolation which contradicts the needs of interactive data exploration. This paper structures, formalizes, and discusses possibilities to enable user involvement in ongoing computations. Based on a structured characterization of needs regarding intermediate feedback and control, the main contribution is a formalization and comparison of strategies for achieving user involvement for algorithms with different characteristics. In the context of integration, we describe considerations for implementing these strategies either as part of the visualization tool or as part of the algorithm, and we identify requirements and guidelines for the design of algorithmic APIs. To assess the practical applicability, we provide a survey of frequently used algorithm implementations within R regarding the fulfillment of these guidelines. While echoing previous calls for analysis modules which support data exploration more directly, we conclude that a range of pragmatic options for enabling user involvement in ongoing computations exists on both the visualization and algorithm side and should be used.

A four-level focus+context approach to interactive visual analysis of temporal features in large scientific data

In this paper we present a new approach to the interactive visual analysis of time‐dependent scientific data – both from measurements as well as from computational simulation – by visualizing a scalar function over time for each of tenthousands or even millions of sample points. In order to cope with overdrawing and cluttering, we introduce a new four‐level method of focus+context visualization. Based on a setting of coordinated, multiple views (with linking and brushing), we integrate three different kinds of focus and also the context in every single view. Per data item we use three values (from the unit interval each) to represent to which degree the data item is part of the respective focus level. We present a color compositing scheme which is capable of expressing all three values in a meaningful way, taking semantics and their relations amongst each other (in the context of our multiple linked view setup) into account. Furthermore, we present additional image‐based postprocessing methods to enhance the visualization of large sets of function graphs, including a texture‐based technique based on line integral convolution (LIC). We also propose advanced brushing techniques which are specific to the time‐dependent nature of the data (in order to brush patterns over time more efficiently). We demonstrate the usefulness of the new approach in the context of medical perfusion data.

Quantifying and Comparing Features in High-Dimensional Datasets

Linking and brushing is a proven approach to analyzing multi-dimensional datasets in the context of multiple coordinated views. Nevertheless, most of the respective visualization techniques only offer qualitative visual results. Many user tasks, however, also require precise quantitative results as, for example, offered by statistical analysis. In succession of the useful Rank-by-Feature Framework, this paper describes a joint visual and statistical approach for guiding the user through a high-dimensional dataset by ranking dimensions (1D case) and pairs of dimensions (2D case) according to statistical summaries. While the original Rank-by-Feature Framework is limited to global features, the most important novelty here is the concept to consider local features, i.e., data subsets defined by brushing in linked views. The ability to compare subsets to other subsets and subsets to the whole dataset in the context of a large number of dimensions significantly extends the benefits of the approach especially in later stages of an exploratory data analysis. A case study illustrates the workflow by analyzing counts of keywords for classifying e-mails as spam or no-spam.

Comparative Visual Analysis of 2D Function Ensembles

In the development process of powertrain systems, 2D function ensembles frequently occur in the context of multi‐run simulations. An analysis has many facets, including distributions of extracted features, comparisons between ensemble members and target functions, and details‐on‐demand. The primary contribution of this paper is a design study of an interactive approach for a comparative visual analysis of 2D function ensembles. The design focuses on a tight integration of domain‐oriented and member‐oriented visualization techniques, and it seeks to preserve the mental model of 2D functions on multiple levels of detail. In this context, we propose a novel focus+context approach for visualizations relying on data‐driven placement which is based on labeling. We also extend work on feature‐preserving downsampling of 2D functions. Our design supports a comparison of 2D functions based on juxtaposition, overlay, and explicit differences. It also enables an analysis in terms of extracted scalar features and 1D aggregations. An evaluation illustrates a workflow in our application context. User feedback indicates a time saving of 70% for common tasks and a qualitative gain for the entire development process.